Bioconjugate Chem. 1997, 8, 664−672
664
Glucose-Induced Release of Glycosylpoly(ethylene glycol) Insulin Bound to a Soluble Conjugate of Concanavalin A Feng Liu,† Soo Chang Song,† Don Mix, Miroslav Baudysˇ, and Sung Wan Kim* Department of Pharmaceutics and Pharmaceutical Chemistry/Center for Controlled Chemical Delivery, Biomedical Polymers Research Building, Room 205, University of Utah, Salt Lake City, Utah 84112. Received February 3, 1997X
Treatment of diabetes mellitus by insulin injections provides long-term control of the disease but lacks any feedback response to glucose concentration changes, which finally leads to a number of life-threatening conditions. The purpose of this study was to improve and optimize an implantable, concanavalin A (Con A) based, glucose-responsive insulin delivery system studied earlier [Jeong, S. Y., Kim, S. W., Holmberg, D. L., and McRea, J. C. (1985) J. Controlled Release 2, 143-152], which can be used for long-term diabetes treatment. To optimize the “insulin component” of the delivery system, we prepared PheB1 insulin amino group monosubstituted monoglucosylpoly(ethylene glycol) (G-PEG) insulin conjugates (PEG Mr 600 or 2000), which showed preserved bioactivity, significantly improved solubility and solution stability at neutral pH, and substantially suppressed hexamerization/ dimerization. To improve the delivery system further, we synthesized and characterized a conjugate of Con A and monomethoxypoly(ethylene glycol) (mPEG, Mr 5000) grafted hydrophilic poly(vinylpyrrolidone-co-acrylic acid) (PVPAA) with Mr of 250 000. The optimal conjugate contained around eight PEG chains and two to three Con A tetramers attached through the amide bonds to the PVPAA chain. The Con A sugar binding characteristics were preserved, and, more importantly, Con A solubility at pH 7.4 substantially increased. This also holds true for a complex formed by the Con A conjugate and G-PEG insulin, which is soluble and does not precipitate under the physiologically relevant conditions under which the complex formed by the Con A conjugate and glycosyl insulin immediately precipitates. Finally, no leakage of the Con A conjugate from a membrane device was detected. Preliminary in vitro release experiments with Con A conjugate and G-PEG insulin complex enclosed in the membrane device showed a pulsative, reversible release pattern for G-PEG insulin in response to glucose challenges of 50-500 mg/dL, demonstrating the feasibility of the release system for use in planned, chronic in vivo studies with diabetic (pancreatectomized) dogs.
INTRODUCTION
The need to improve the control of glucose homeostasis and to avoid frequent chronic hyperinsulinemic episodes caused by conventional parenteral insulin injections in the treatment of diabetes led in the late 1970s to the development of different continuous insulin delivery systems (1). However, most of these systems suffered from the lack of feedback regulation utilizing some kind of glucose level monitoring in vivo; therefore, new approaches have been used to develop glucose-responsive insulin delivery systems (2-8). Our group has been working for a long time on a unique delivery system based on competitive binding (9), namely the displacement of glycosylated insulin by glucose from concanavalin * Author to whom correspondence should be addressed [telephone (801) 581-6654; fax (801) 581-7848; e-mail rburns@pharm. utah.edu]. † Both contributed equally to the project. X Abstract published in Advance ACS Abstracts, August 15, 1997. 1 Abbreviations: Con A, concanavalin A; PEG, poly(ethylene glycol); G-PEG, glycosyl(glucosyl)poly(ethylene glycol); mPEG, monomethoxypoly(ethylene glycol); mPEG-NH2, aminomethoxypoly(ethylene glycol); PVPAA, poly(vinyl pyrrolidone-co-acrylic acid); mPEG-PVPAA, mPEG-grafted PVPAA; mPEG-PVPAA-Con A, conjugate of Con A and mPEG-PVPAA; Hepes, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid; AcOH, acetic acid; TFA, trifluoroacetic acid; TBA, tributylamine; TEA, triethylamine; IBCF, isobutylchloroformate; Boc, tert-butyloxycarbonyl; DMF, dimethylformamide; DMSO, dimethyl sulfoxide; MWCO, molecular weight cuttoff; AUC, area under the curve; SEC, size exclusion chromatography.
S1043-1802(97)00128-6 CCC: $14.00
A (Con A).1 The system is enclosed in a polymeric membrane device that is permeable to glucose and glycosylated insulin and nonpermeable to Con A. The current version of the system is a rechargeable pouch that can be implanted into the peritoneal cavity and then be refilled from the outside (10). Nevertheless, the performance of such a device is hampered by the limited solubility of glycosylated insulin-Con A complex, which causes a long unwanted lag phase for the onset of release of glycosylated insulin derivative when the device is challenged by high glucose concentration (11). The physicochemical basis for this phenomenon is the tetrameric nature of the Con A molecule (12) and the limited, yet partially preserved, ability of glycosylated insulin derivatives to form dimers and hexamers (13). This especially holds true for GlyA1 substituted derivatives that comprise the major component in glycosylated insulin preparations used in the release studies so far (14). Recently, we worked out a new synthetic scheme for glycosylated insulins enabling us to attach a sugar moiety specifically to the PheB1 amino group (15). Generally, this site-specific modification does not have a negative impact on the binding of such derivatives to insulin receptors, and thus their biological activity is more or less preserved (16). We have also shown that these glycosylated derivatives are predominantly dimeric over the considered concentration range (15). We hypothesized that by placing a sufficiently long poly(ethylene glycol) (PEG) spacer between the sugar moiety and the above specified amino group of the insulin molecule (Scheme 1), we would achieve further reduction in the © 1997 American Chemical Society
Glycosyl-PEG Insulin and Con A Conjugate for Insulin Delivery Scheme 1. Preparation of PheB1-Substituted Monoglycosyl Poly(ethylene glycol) Insulin
dimerization of glycosylpoly(ethylene glycol) insulin (GPEG insulin), thus preventing precipitation of its complex with Con A and improving its release profile from a pouch. Moreover, the attachment of PEG chain(s) to the insulin molecule should have a beneficial impact on several important characteristics of the hormone, such as diminished immunogenicity and antigenicity, as has already been demonstrated for PEG insulin conjugates (17), reduced susceptibility to proteolysis, and increased plasma half-life as well as significantly improved physical stability, as has been shown for many pharmaceutically relevant proteins (18). The last feature is especially important for the pouch design since insulin solutions at elevated temperature (37 °C) and constant irregular motions (hydrodynamic stress) tend to form insoluble fibrils with negligible bioactivity (19-22). To further optimize physicochemical and biological features of the glucose-responsive insulin delivery system described above for in vivo use, unsolved problems associated with Con A had to be also addressed. Originally, unmodified Con A was used, which was slowly leaking from a pouch (23). To prevent the leakage, new insoluble Con A systems were designed based on Con A immobilized onto a solid support (24) or Con A crosslinked as microspheres (14), which had the disadvantage that the beads or microspheres settled to the bottom of the pouch, preventing reliable and reproducible insulin release during glucose challenge experiments. Thus, soluble Con A oligomer (Mr > 300 000) was prepared by glutaraldehyde cross-linking (25). Still, reproducibility problems persisted because of limited solubility of Con
Bioconjugate Chem., Vol. 8, No. 5, 1997 665 Scheme 2. Preparation of mPEG-NH2 Grafted Poly(vinylpyrrolidone-co-acrylic acid) and Conjugation of Concanavalin A
A oligomer and the insolubility of Con A oligomerglycosylated insulin complex. To increase Con A solubility and prevent Con A leakage from the pouch as well as precipitation of the complex, we decided to synthesize a conjugate of Con A and hydrophilic copolymer poly(vinylpyrrolidone-co-acrylic acid) (PVPAA) with high Mr of 250 000 (Scheme 2). Grafting of mPEG-NH2 molecules to PVPAA before Con A conjugation should further increase the solubility of Con A and its complex with glycosylated insulin and especially with G-PEG insulin. This paper describes the synthesis, purification, and chemical, physical, and biological characterization of G-PEG insulin derivatives with various lengths of PEG spacer. Preparation and characterization of a high Mr and soluble Con A conjugate with mPEG-grafted PVPAA (mPEG-PVPAA) are also described. Finally, the release behavior of G-PEG insulin from mPEG-PVPAA-Con A conjugate enclosed within a porous polymer membrane device at various glucose concentrations was studied to demonstrate the feasibility of using this system for selfregulating insulin delivery in vivo. EXPERIMENTAL PROCEDURES
Materials. Poly(ethylene glycol) diacid (PEG diacid, Mr 600 and 2000) was purchased from Fluka (Ronkonkoma, NY), and aminomethoxypoly(ethylene glycol) (mPEGNH2) of Mr 5000 was obtained from Shearwater Polymers (Huntsville, LA) and dried over P2O5 under vacuum before use. Human crystalline Zn-insulin was obtained from Bayer Corp. (Kankakee, IL). p-Aminophenyl R-Dglucopyranoside, methyl R-D-mannopyranoside, dansyl chloride, ammonium bicarbonate, anisole, urea, Trizma base, Hepes, and concanavalin A-Sepharose were purchased from Sigma Chemical Co. (St. Louis, MO). Concanavalin A (Sigma) was purified by precipitation of the fragmented fraction in ammonium bicarbonate (12).
666 Bioconjugate Chem., Vol. 8, No. 5, 1997
Acetic (AcOH), trifluoroacetic (TFA), and sulfuric acids, isobutylchloroformate (IBCF), tributylamine (TBA), triethylamine (TEA), di-tert-butyl dicarbonate (Boc dicarbonate), and all other reagents as well as all organic solvents used were at least ACS grade and purchased from Aldrich Chemical Co. (Milwaukee, WI). Spectra/ Por dialysis tubing (MWCO 1000 and 3500) and Spectra/ Por RC DispoDialyzer (1 mL, MWCO 10 000, 25 000, 50 000) were purchased from Spectrum Medical Industries (Houston, TX). Ultrafiltration membranes (PCMK regenerated cellulose disks) with MWCO 300 000 were from Millipore (Bedford, MA). Poly(vinylpyrrolidone-coacrylic acid) (PVPAA) (Mr 250 000, VP/AA molar ratio ) 75/25) was obtained from ISP Technologies, Inc. Insulin radioimmunoassay kits from ICN Micromedics, Horsham, PA, were used. Synthesis of Glucosyl-PEG Diacid (G-PEG Diacid). The mixed anhydride method of amide bond formation was used to couple one of the carboxylic groups of PEG diacid to the amino group of glucoside (26). A 1 g (1.67 mmol) portion of PEG diacid (Mr 600) containing 3.34 mmol of free carboxyl groups was dissolved in 3 mL of dry dimethylformamide (DMF) containing 1.83 mmol of TBA in an ice bath under N2. IBCF (1.83 mmol) dissolved in 1 mL of dry DMF was added and the mixture stirred for 15 min. p-Aminophenyl R-D-glucoside (1.67 mmol) dissolved in 3 mL of dry DMF was added dropwise, and the mixture was stirred for an additional 15 min in an ice bath and for 3 h at room temperature. The product was precipitated in an excess of dry ether (200 mL). The precipitate was dissolved in 20 mL of distilled water, dialyzed against distilled water (tubing with MWCO 1000), and freeze-dried. The unreacted PEG diacid was removed from the mixture on a Con A-Sepharose column under conditions described later (see G-PEG Insulin Synthesis). The fraction eluted by 0.1 M methyl R-Dmannopyranoside was collected, dialyzed extensively against water, and freeze-dried. Alternatively, PEG diacid (Mr 2000) was used and an identical procedure followed. Synthesis and Purification of PheB1-Substituted G-PEG Insulins. Di-NR-Boc-GlyA1, N-Boc-LysB29 insulin was synthesized and purified as described earlier (27). DiBoc-GlyA1, LysB29 insulin derivative (26 µmol) was dissolved in 6 mL of dry dimethyl sulfoxide (DMSO) and activated G-PEG diacid (Mr 600) added. To prepare activated G-PEG diacid, 173 mg of the G-PEG diacid mixture containing 130 µmol of free carboxyl groups was dissolved in 3 mL of dry DMF containing 130 µmol of TBA and 130 µmol of IBCF and stirred for 15 min under N2 in an ice bath. Finally, the reaction mixture was stirred for 3 h at room temperature under N2 and the reaction stopped by the addition of excess of aminoethanol (1.3 mmol). The resulting insulin derivative was precipitated in excess of dry acetone (100 mL) containing 65 µL of 6 N HCl. The precipitate was washed with 10 mL of ethanol twice, dissolved, dialyzed against 0.01% NH4HCO3, and finally lyophilized. This material was dissolved in 5 mL of anhydrous TFA with 5% anisole as a scavenger and kept under N2 at 0 °C for 1 h to remove Boc groups. Deprotected insulin derivative was precipitated in an excess of dry ether, dissolved, dialyzed against 0.01% NH4HCO3, and lyophilized. Alternatively, PEG diacid (Mr 2000) was used and an identical procedure followed. Purification of G-PEG insulin derivatives (PEG 600, PEG 2000) was achieved first using FPLC on a Mono S preparative column (Pharmacia AB) using 1 M AcOH and 7 M urea as eluent and a gradient of NaCl as described earlier (13, 15). In each case, only the major peak was
Liu et al.
collected, dialyzed, and lyophilized. Final purification was achieved by affinity chromatography on a Con A-Sepharose column. The lyophilisate (40 mg) was dissolved in 40 mL of 0.02 M Tris buffer, pH 7.4, containing 0.5 M NaCl, 3 mM CaCl2, and 3 mM MnCl2 and loaded on a Con A-Sepharose column (40 mL) at a flow rate 0.15 mL/min. The column was eluted until absorbance at 280 nm (A280) dropped below 0.01 value, and then identical buffer containing 0.1 M methyl R-Dmannopyranoside was applied. The specifically eluted peak of G-PEG insulin derivative was detected by monitoring A280, collected, dialyzed against water and 0.01% NH4HCO3, and finally lyophilized. Characterization of G-PEG Insulins. To assess physical stability, G-PEG insulin derivatives were dissolved in 0.01 M PBS (0.15 M NaCl, 0.01% sodium azide), pH 7.4, at 80 µM concentration and filtered through a 0.22 µm filter. Zn-Insulin, as well as Zn-free insulin solutions prepared in the same manner were used as controls. The aggregation was evaluated in 5 mL of borosilicate glass vials (1.2 mL filling volume) at 37 °C, 100 strokes/min. Individual vials (n ) 12) were periodically withdrawn at preselected periods of time, and the remaining soluble fraction of insulin derivative was determined after filtration (PVDF membrane, 0.22 µm) using ultraviolet spectroscopy or HPLC. The biological activity of insulin derivatives was tested in a rat model (male Sprague-Dawley rats, 250 ( 50 g) by measuring blood glucose depression levels. All procedures in handling animals adhered to the “Principles of Laboratory Animal Care” (NIH Publication 85-23, revised 1985). The animals were fasted overnight (16 h) prior to the experiment. In the morning, the rats were anesthetized with sodium pentobarbital intraperitoneally. Each animal received an intravenous (iv) injection (tail vein) of the particular insulin derivative [3.4 nmol (mL of NS)-1 /kg-1]. Blood samples were taken from the jugular vein 15 and 5 min before injection and 15, 30, 60, 90, 120, 180, and 240 min after injection. Blood glucose levels were measured with an Accucheck III blood glucose monitor (Boehringer Mannheim). Bioactivities of insulin derivatives were calculated from area under the curve (AUC) values and the AUC value for the same dose of unmodified human insulin. Synthesis of Comb-Type Graft Copolymer of PVPAA and mPEG-NH2. One gram of PVPAA was dissolved in 20 mL of dry DMF containing 0.48 mL of TEA in an ice-water bath. The solution was purged with dry nitrogen for 10 min. IBCF (18, 36, or 72 µmol, respectively) was added to the above solution with stirring. After 5 min, 90, 180, or 360 mg (18, 36, or 72 µmol) of mPEG-NH2 respectively, dissolved in 5 mL of dry DMF was added dropwise. The mixture was kept at 0 °C for 30 min and at room temperature for an additional 30 min. The reaction solution was then poured into 500 mL of ethyl ether, and the precipitate was collected and washed with ethyl ether. mPEG-PVPAA was purified by ultrafiltration using membrane with MWCO of 300 000 until no mPEG-NH2 was detected in the ultrafiltrate by the ninhydrin method (28). Finally, the product was dialyzed against distilled water and lyophilized. The amount of grafted mPEG in the copolymer was analyzed by 1H-NMR spectroscopy (200 MHz, D2O) through the integration of PEG polymer signal (δ 3.62). Synthesis and Characterization of mPEG-PVPAA-Con A Conjugate. mPEG-PVPAA copolymer (50 mg) was dissolved in 1.0 mL of DMSO containing 140 µL of TEA and the solution purged with nitrogen. IBCF (1 µmol) was added and the mixture stirred for 5 min at room temperature. The activated polymer solution was
Glycosyl-PEG Insulin and Con A Conjugate for Insulin Delivery
then added to 20 mL of Hepes buffer (0.05 M, pH 7.4, 3 mM CaCl2, 3 mM MnCl2, and 0.9% NaCl) containing 100 mg of Con A at 0 °C. The reaction was allowed to proceed for 30 min at 0 °C and then for 30 min at ambient temperature. The product was purified by ultrafiltration using membrane with MWCO of 300 000 until no A280 was detected in the ultrafiltrate. Traces of turbidity were removed by centrifugation (5000 rpm, 4 °C, 30 min). The Con A content in the final concentrated solution of the conjugate was assayed by A280 (25). To compare the binding characteristics of conjugated Con A with that of free Con A, a glycogen precipitation assay was employed. Glycogen in the Hepes buffer (see above, 100 µL) was added to a 1 mL stirred UV cuvette containing mPEG-PVPAA-Con A or Con A in the same buffer (900 µL). The final concentrations of Con A (in both free and conjugated forms) and glycogen were 200 and 60 µg/mL, respectively. The increase in turbidity was monitored continuously through absorbance changes at 480 nm (29). Solubility of complexes formed by mPEG-PVPAACon A or free Con A and glycosyl insulin or G-PEG insulin was also investigated. Known amounts of Con A or mPEG-PVPAA-Con A and insulin derivatives in Hepes buffer were mixed in a 1 mL UV cuvette and incubated at 37 °C for 10 min. The concentration of free Con A or Con A in the polymer conjugate was 1 mg/mL; the concentration of insulin derivatives varied from 4 to 40 µM. The solubility of complexes was assayed by turbidity measurements at 480 nm. Preliminary in Vitro G-PEG Insulin Release Experiments. One milliliter of Hepes buffer containing mPEG-PVPAA-Con A conjugate (4 mg/mL of Con A) and appropriate insulin derivative (500 µg/mL) was transferred into a DispoDialyzer with a sample volume of 1 mL (membrane MWCO 25 000). The loaded microdialyzer was placed into a small column (inner volume of 6 mL) equipped with a water jacket. The temperature was maintained at 37 °C. The column with microdialyzer was washed upward at a constant flow rate of 9 mL/h with Hepes buffer containing glucose. The glucose concentration changed every 4 h between 50 and 500 mg/ dL. Sixty minute fractions of the eluent were collected and assayed for G-PEG insulin using RIA commercial kit (Coat-A-Count) according to the manufacturer’s instructions. An RIA standard curve was constructed using G-PEG insulin samples of known concentrations. Analytical Methods. Quantitative analysis of sugar residue (hexose type) content in the conjugates was performed by using the phenol/sulfuric acid method (30). p-Aminophenyl R-D-glucopyranoside was used to construct calibration curves. Alternatively, a UV spectrophotometric assay was used taking advantage of the strong chromophoric acylamidophenoxy group presence in the PEG conjugates. p-Succinamidophenyl R-D-glucopyranoside (15) was used as a standard to the construct calibration curve at 244 nm (λmax ) 1.02 × 104 M-1 cm-1). Free N-terminal amino acids in insulin derivatives were determined by the N-terminal dansylation technique. Released dansyl amino acids were identified after acid hydrolysis (6 N HCl, 105 °C, 8 h) by two-dimensional TLC (31). UV absorption spectra were obtained on a PerkinElmer UV-vis Lambda 19 spectrophotometer. CD absorption spectra for G-PEG insulin derivatives were obtained by the use of a JASCO J-720 spectropolarimeter. In this case, samples (0.5 mg/mL or lower) were dissolved in phosphate-buffered saline (0.01 M phosphate; 0.15 M NaCl) (PBS), pH 7.4, containing 0.01% azide, and the
Bioconjugate Chem., Vol. 8, No. 5, 1997 667
ellipticity of the solution was measured in a cylindrical quartz cell (0.1 cm optical path length). Analytical Chromatographic Procedures. The concentration dependence of the elution volume for G-PEG insulins was examined by size exclusion chromatography (SEC) on a Superose 12 (10 mm × 30 cm) column (Pharmacia AB) using FPLC. The column was eluted with 0.05 M Tris/HCl, pH 8.0, at a flow rate 0.4 mL/min at 22 °C. A 0.2 mL sample was injected, and the eluent was continuously monitored at 280 nm. The column was calibrated with molecular weight standards to generate a calibration curve (log Mr vs Ve/Vo) to enable calculation of the apparent Mr of insulin derivatives. Homogeneity and concentration of insulin derivatives in solution were determined using reversed phase HPLC (Waters modular system with Waters Model 745 integrator). The Vydac C4 column (4.6 mm × 25 cm) was equilibrated with 0.1% TFA and 20% acetonitrile (solvent A) at a flow rate of 1 mL/min. A 100 µL volume of each sample was injected, and a linear gradient of solvent B (0.1% TFA in 90% acetonitrile) with a slope of 2% B/min was applied. A280 of the eluent was recorded and processed to evaluate insulin derivative content. Alternatively, homogeneity was examined on analytical Mono S HR 5/5 column (Pharmacia AB) equilibrated with 1 M AcOH, 7 M urea, and 0.01 M NaCl under conditions described earlier (15). RESULTS AND DISCUSSION
G-PEG Diacid Synthesis and Characterization. To achieve specific binding of PEG insulin derivative to concanavalin A, an aminophenyl glucopyranosyl moiety had to be first attached to PEG diacid (Scheme 1). A mixed anhydride method to activate carboxylic groups of PEG diacid was used. The molar chloroformate/PEG diacid ratio used was kept slightly over 1.0 (exactly 1.1: 1) because of the traces of water from PEG diacid preparation that were very difficult to remove. These conditions resulted in quantitative coupling and preferential formation of monoglucosyl-PEG (monoG-PEG) diacid conjugate. As expected according to the statistical nature of the reaction, however, diglucosyl-PEG (diGPEG) diacid conjugate, as well as unreacted PEG diacid, was also present in the mixture. The molar ratio of the three components comprising the mixture was 2 (monoGPEG diacid):1 (diG-PEG diacid):1 (PEG diacid). This was indirectly shown by volumetric titration analysis, which revealed that 51 ( 3 mol % of carboxylic groups were no longer available, and by glucose content determination (phenol/sulfuric acid destruction method and UV spectrophotometric method; see Experimental Procedures), which was 1.05 ( 0.09 mol/mol of conjugate. In both cases, the Mr of the conjugate was considered to be equal to the Mr of monoG-PEG diacid. Unreacted PEG diacid was removed from the mixture by affinity chromatography on a Con A-Sepharose column (unbound fraction). The specifically eluted fraction was a mixture of monoG-PEG diacid and diG-PEG diacid conjugates at a molar ratio 2:1, again as determined through glucose residues quantitation (1.3 mol of glucose/mol of PEG diacid) and titration of carboxylic groups (0.7 mol of COOH/mol of PEG diacid). The presence of disubstituted PEG diacid derivative does not interfere with insulin modification, and unreacted PEG diacid, the presence of which would result in formation of covalent insulin dimers (through PEG chain spacer), has been removed. Thus, the Con A-Sepharose-purified G-PEG diacid fraction was used directly for insulin modification. The overall yield was 30%.
668 Bioconjugate Chem., Vol. 8, No. 5, 1997
Liu et al.
Figure 1. Analytical cation exchange chromatography of G-PEG(600) insulin on a Mono S HR 5/5 column (Pharmacia AB). The column was equilibrated with buffer A (1 M AcOH, 7 M urea, 0.01 M NaCl) at a flow rate 1 mL/min. Sample (200 µL, 1 mg/mL buffer A) was injected and the column eluted with a linear gradient of buffer B (1 M AcOH, 7 M urea, 0.3 M NaCl) with a slope 2% B/min.
Figure 2. Analytical reversed phase HPLC of G-PEG(600) insulin on a Vydac C4 column (Separations Group). The column was equilibrated with buffer A (0.1% TFA, 20% acetonitrile) at a flow rate 1 mL/min. Sample (100 µL, 1 mg/mL solvent A) was injected and the column washed isocratically for 5 min. A linear gradient of solvent B (0.1% TFA, 90% acetonitrile) with a slope 2% B/min was then applied.
PheB1-Substituted G-PEG Insulin Synthesis and Purification. It is a well-documented fact that the reactivity of the PheB1 amino group of insulin toward an electrophilic substitution group is low compared to that of the GlyA1 and LysB29 amino groups (32). We previously overcame this problem by the reversible protection of GyA1 and LysB29 amino groups by a Boc group and synthesizing diBoc-GlyA1, LysB29 insulin derivative. Using this intermediate (Scheme 1), specific modification reaction on the PheB1 amino group can be accomplished. Specifically, a 5 molar excess of activated monoG-PEG diacid (Mr 600 or 2000) over diBoc-insulin (taking into account 33 mol % contamination with diGPEG diacid, see above) had to be used to achieve quantitative modification of the PheB1 amino group. After the removal of Boc groups with TFA, PheB1substituted G-PEG insulin was purified by two-step chromatography on Mono S and concanavalin A columns to separate residual unreacted G-PEG derivatives and other impurities. The 5 molar excess of activated monoGPEG diacid could result not only in the specific acylation of the only available insulin PheB1 amino group but also in side reactions, specifically O-acylation of insulin molecule through insulin hydroxyl groups. However, if any esters of monoG-PEG diacid and insulin were formed, they were later hydrolyzed by very extensive dialysis in ammonium bicarbonate under basic conditions. The overall yield was about 25%, and homogeneity of synthesized G-PEG insulins was established by analytical ion exchange (Figure 1) and reversed phase chromatography (Figure 2). Characterization of G-PEG Insulins. The PheB1 amino group of insulin as the site of attachment of G-PEG moiety was confirmed by N-terminal group analysis. Only dansylglycine and dansyl--lysine, which corresponds to a single inner-chain lysine residue, were detected by twodimensional TLC. No dansylphenylalanine was found (Table 1). Glucose content also pointed to the presence of exactly one G-PEG moiety in both conjugates. Moreover, the retention time of G-PEG insulins on an analytical Mono S column coincided exactly with that of other monosubstituted (monoamidated) insulin derivatives (15). To have further proof that only one G-PEG moiety was attached to the insulin molecule and, thus, no G-PEG
Table 1. Characterization of Glucosyl-PEG Insulins
derivative
N-termin- glucose biol act.a fibrillation ( SD period ( SD al amino content (days) acid (mol/mol) (IU/mg)
insulin Gly, Phe G-PEG(600)b insulin Gly G-PEG(2000)b insulin Gly
0.05 1.05 0.89
25.2 ( 4.7 21.2 ( 4.2 15.2 ( 3.6
0.5 ( 0.2 26.4 ( 5.4 32.3 ( 4.3
a The biological activity of each derivative was determined with at least six animals. b The site of G-PEG attachment is PheB1 amino group.
Figure 3. UV spectra of Zn-free insulin (I), G-PEG(600) insulin (II), and PheB1-LysB29-diglucosyl insulin (III) at concentration 68 µM. UV spectra of G-PEG(2000) insulin and PheB1-monoglucosyl insulin overlapped with that of G-PEG(600) derivative and therefore have not been included.
moieties were coupled to the insulin molecule through O-acylation, UV spectra of G-PEG insulins were examined. The UV spectrum of G-PEG(600) insulin is shown in Figure 3 and compared to the UV spectrum of Zn-free insulin and PheB1-LysB29-diglucosyl insulin that contains two succinamidophenyl R-D-glucopyranoside moieties coupled to the insulin molecule (15). Not shown, for visual clarity, are the UV spectra of PheB1-monoglucosyl insulin, containing one succinamidophenyl R-Dglucopyranoside group, and G-PEG(2000) insulin, which both coincided with the UV spectrum of G-PEG(600) insulin. It is evident that the region between 240 and 260 nm is very sensitive to the number of succinamidophenyl R-D-glucopyranoside groups coupled to the insulin molecule. Thus, identical spectra of PheB1-G-
Bioconjugate Chem., Vol. 8, No. 5, 1997 669
Glycosyl-PEG Insulin and Con A Conjugate for Insulin Delivery
Figure 4. Effect of insulin and G-PEG(600) insulin on blood glucose levels in vivo. The peptides were administered to fasted rats by intravenous injection of a dose of 3.4 nmol/kg. Vertical bars represent standard deviation (n g 6). See Experimental Procedures for further details.
PEG insulins and PheB1-monoglucosyl insulin, together with a failure to detect any trace of dansyl-Phe, can be taken as a direct proof that only one G-PEG moiety is attached specifically to the insulin PheB1 amino group. The bioactivity of the conjugates was tested in vivo by blood glucose depression test in rats using an iv dose of 3.4 nmol kg-1 mL-1 of a specific derivative (Table 1; Figure 4). It can be concluded that the bioactivity of the G-PEG insulin derivative with lower Mr of attached PEG(600) has slightly lower bioactivity than insulin, but this difference is not statistically significant (Student’s t-test, p > 0.2), which agrees with the findings of Neubauer et al. (33). As the Mr of attached PEG is increased (2000), the bioactivity decreases further and the difference becomes significant (p < 0.002). To exclude the possibility that changes in secondary/tertiary structure of the insulin molecule occurred because of modification, CD spectra of G-PEG insulins were compared to that of Znfree insulin (Figure 5). The spectra practically overlap, indicating that no spatial structural changes occurred. Previously, PheB1-substituted high Mr PEG (5500) insulin conjugate was synthesized and found to be practically devoid of any biological activity (34). Its CD spectrum, contrary to our findings, was substantially changed. It has been well established that PheB1 and its vicinity are not involved in the binding to insulin receptor (16, 35). Thus, in our case, the negative effect of PEG moiety attached to PheB1 on bioactivity of G-PEG insulins can be most likely explained by nonspecific sterical hindrance to the receptor binding by the PEG moiety, caused by its large hydrodynamic volume (36). Also, as the attached PEG moiety Mr is increasing further and becoming comparable to insulin (34), it has a direct negative effect on the insulin conformation, thus decreasing the conjugate bioactivity further. The long-term physical stability of PheB1-substituted G-PEG insulins was investigated by assessing their ability to form fibrils in an accelerated shaking test (19, 22). Insulin aggregation constitutes a major problem in a long-term steady insulin administration by insulin pumps (1). Solutions of G-PEG insulins (80 µM) were vigorously shaken at 37 °C for extended periods of time. Samples were periodically withdrawn and filtered, and the residual concentration of insulin derivative was
Figure 5. Far-UV CD spectra of Zn-free insulin and G-PEG(600) insulin in PBS, pH 7.4, at a concentration of 75 µM. Mean residual ellipticity (θ)λ was calculated using the expression (θ)λ ) θλM0/104Cl, where θλ is the observed ellipticity at wavelength λ, M0 is the mean residue molecular weight, which for G-PEG(600) insulin was calculated to be 132, C is the protein concentration in g/mL, and l is the optical path length in cm. The CD spectrum of G-PEG(2000) insulin overlapped with that of G-PEG(600) derivative and therefore has not been included. Table 2. Concentration Dependence of Association State of G-PEG Insulins As Determined by SECa association state (Mr app/Mr monomer) derivative 2Zn-insulin G-PEG(600) insulin G-PEG(2000) insulin a
c) c) c) c) 0.17 mM 0.43 mM 0.87 mM 1.73 mM 2.0 1.2 1.1
3.4 1.6 1.6
4.2 1.9 2.2
4.6 2.2 2.3
On Superose 12 column.
measured by UV or HPLC. In parallel, CD spectra for these samples were recorded to ensure that no conformational changes took place in the soluble fractions of the samples. Zn-insulin and Zn-free insulin were used as controls (Table 1). G-PEG insulins showed excellent physical stability, even 2-3 times higher than that of PheB1-substituted glycosylated insulins (15). Also, some effect of PEG spacer length was observed, even though the difference was not significant. Since insulin forms dimers and hexamers depending on its concentration, association properties of G-PEG insulins were evaluated using SEC on a Superose 12 column (Table 2). It is evident that attachment of PEG derivative chains to PheB1 has the same effect as was observed in the case of glycosylated insulins (13, 15), significantly suppressing hexamer formation without having a large impact on dimer formation. This can be taken as an independent indication that the spatial structure of G-PEG insulin derivatives is preserved. It also indicates, however, that the G-PEG coupling to PheB1 insulin amino group did not sufficiently suppress the dimerization ability of these insulin derivatives. Indeed, when G-PEG(600) insulin solution (1 mg/mL) was intermixed with Con A solution (4 mg/mL), precipitate formed instantly (see further). In general, G-PEG insulin with the smaller Mr PEG spacer of 600 (hereafter, G-PEG insulin) seemed to be suitable for use in a self-regulated insulin delivery system based on concanavalin A, because of its preserved bio-
670 Bioconjugate Chem., Vol. 8, No. 5, 1997
Figure 6. Time course of turbidity formation for free Con A and mPEG-PVPAA-Con A conjugate in glycogen precipitation test. Concentration of free Con A or Con A in the conjugate was 200 µg/mL, and concentration of glycogen was 60 µg/mL in 0.05 M Hepes buffer, pH 7.4, containing 3 mM CaCl2 and MnCl2 and 0.9% NaCl.
activity and increased solution stability, and was used exclusively in further studies. However, its partially preserved dimerization ability causing precipitation of G-PEG insulin-Con A complex still hampered its use in this specific delivery system (see below). Therefore, our efforts then concentrated on improving the solubility of Con A, and especially its complex with G-PEG insulin, through Con A covalent conjugation to mPEG-grafted hydrophilic polymer PVPAA. mPEG-PVPAA-Con A Conjugate Synthesis and Characterization. At first, we intended to synthesize a simple conjugate of Con A and PVPAA to prevent Con A leakage from the pouch and increase solubility of Con A and especially of Con A-G-PEG insulin complex. It turned out, however, that this was not feasible because the high degree of cross-linking resulted in an insoluble conjugate (data not shown). Thus, to limit or prevent cross-linking, we decided to graft mPEG-NH2 chains onto a PVPAA backbone (Scheme 2). Three different molar ratios of mPEG-NH2 to PVPAA, 17:4, 34:4, and 68:4, were used, which resulted in grafting 4.2, 7.8, and 16.1 PEG chains to one molecule of PVPAA, respectively, as demonstrated by 1H NMR spectroscopy. Using these grafted copolymers for Con A conjugation (see Experimental Procedures for details), we discovered that the low-degree PEGylated PVPAA still gave rise to a cross-linked, mostly insoluble, material, while the high-degree PEGylated PVPAA reacted with Con A with very low efficiency. Only mPEG-PVPAA graft copolymer with a medium degree of substitution (7.8 mol of PEG/mol of PVPAA) gave an mPEG-PVPAA-Con A complex that was soluble. This complex contained about 2.5 mol of Con A tetramer per mole of mPEG-PVPAA copolymer. No leakage of Con A or the conjugate was detected from Spectra/Por DispoDialyzers (with different MWCO of up to 50 000) over a 1 week period, as assayed by UV spectroscopy of surrounding dialyzate (5 mL) under static conditions (no flow, see further). This conjugate was used in further studies. The rate of appearance of turbidity resulting from the reaction of glycogen and Con A reflects the state of saccharide-binding activity of Con A in mPEG-PVPAACon A conjugate. As shown in Figure 6, the time courses of precipitation with glycogen for mPEG-PVPAA-Con
Liu et al.
Figure 7. Solubility of complexes formed by free Con A or mPEG-PVPAA-Con A conjugate and glucosyl insulin or GPEG(600) insulin. Concentration of Con A or Con A in the conjugate was 1 mg/mL in 0.05 M Hepes buffer, pH 7.4. The concentration of insulin derivatives was varied, and turbidity formed was measured by absorbance increase at 480 nm, 10 min after mixing.
A conjugate or free Con A do not differ significantly, which indicates that the coupling reaction of Con A and mPEG-PVPAA did not occur within the saccharide binding sites of Con A. The glycogen turbidity curve (see Figure 6 for conditions) after 1 week of preincubation of the conjugate in the precipitation buffer (Hepes buffer, pH 7.4, 37 °C) was identical to the turbidity curve obtained with mPEG-PVPAA-Con A conjugate freshly dissolved in the precipitation buffer and incubated with glycogen, indicating that the Con A binding activity in the conjugate is preserved for at least 7 days. Having established that the binding characteristics of Con A in the conjugate are preserved, we investigated the solubility of complexes formed by mPEG-PVPAACon A and glycosyl-PheB1-insulin (15) or G-PEG insulin as a function of the insulin derivatives’ concentrations (Figure 7). For comparison, turbidity measurements for free Con A and glycosyl insulin or G-PEG insulin complexes are also included. It is evident that the solubility of the complex formed by mPEG-PVPAA-Con A conjugate and G-PEG insulin is superior to those of other complexes formed, even though for high G-PEG insulin concentration a very slight turbidity was detected (Figure 7). Insulin replacement therapy usually requires administration of 1-2 mg of insulin per day per patient. Thus, to be closer to therapeutically relevant conditions, we also investigated long-term solubility/stability of mPEG-PVPAA-Con A conjugate and G-PEG insulin mixture at concentrations 4 and 0.5 mg/mL, respectively. No precipitate was detected at 37 °C for at least 7 days. In contrast, if glycosyl insulin at the same concentration was used, macroscopic precipitate formed sediment in
